At present, to save deployment, operation and maintenance cost and human cost, many Internet companies generally choose to lease a cloud server cluster instead of building their own computer rooms, and use the leased cloud server cluster to build desired systems such as a mail service system, an E-commerce system, a big data computing system.
As for the system deployment, the current practice usually deploys different systems to the same cloud server cluster, or different systems to different cloud server clusters. Although the current deployment methods enable the systems as built to operate normally, they also have following defects: deploying different systems to the same cloud server cluster likely causes competition for resources such as CPU (Central Processing Unit) and memory, and very likely affects the stable operation of the systems; deploying different systems to different cloud server clusters requires the lease of the corresponding cloud server cluster based on the maximum access amount of the respective systems. However, the access amount of some systems has a notable on-peak/off-peak phenomenon, for example, service systems such as E-commerce system and mail system have very few visits in the early morning, when it is quite likely the cloud server resources are wasted.